Ornith-1.5: Self-Improving Open-Source MoE Models Rival Claude Opus
aftahi_ai · x · 2026-08-19
Ornith-1.5 is a family of open-source LLMs (9B Dense, 35B MoE, 397B MoE) released under the MIT license. It trains via end-to-end self-improvement strategies, enabling the model to propose tasks, generate scaffolds, and use experiences for reinforcement learning.
It achieves state-of-the-art performance among comparable open-source models:
- Terminal-Bench 2.1: 86.1
- SWE-Bench (verified): 86
- DeepSWE: 56
- HLE: 44.6
Its performance in reasoning, agentic, and coding tasks is claimed to be comparable to Claude Opus 4.8.
Related event: Ornith-1.5 Release: 397B MoE Self-Improvement Targets Claude Opus(24 posts)→
More from coding & agent
- Six-month prediction: hiring for generic agent harnesses built from scratch will cool — yi111 · 2026-10-16
- Omarchy ISO wouldn't boot on NVIDIA, so he had Claude patch it — indragie · 2026-10-16
- Dev uses Grok to clean up 3D printer firmware, catches it sneaking in nonsense code — yacineMTB · 2026-10-16
- A rude AI agent took 89% of a shared resource in 102-run multi-agent personality experiment — Brave-History-6502 · 2026-10-16
- Guardrails can make risky delegation work — but it takes a full-time team, says iandanforth — iandanforth · 2026-10-16
- What can you truly delegate? Only tasks with unambiguous, verifiable results — iandanforth · 2026-10-16